Arxiv:2010.02668V2 [Math.NT] 18 Oct 2020

AN IDENTITY INVOLVING THE CYCLOTOMIC POLYNOMIALS

PABLO L. DE NAPOLI´

Abstract. We present an elementary identity for the cyclotomic polynomials Φn(x) which reﬂects a kind of multiplicative property of Φn(x) as a function of n, and we explore its connections with the properties of other arithmetical functions.

Important Note: In the ﬁrst version of this article uploaded to the arXiv, it is said that this result seemed to be new. How- ever, after that, I have learned that this identity in Theorem 1.1 has previously appeared in [CMW95, corollary 2] (with a diﬀerent proof).

1. Introduction and main result

For each natural number n ∈ N, let Φn(X) denote the n-th cyclotomic polynomial, i.e.: the monic polynomial whose roots are the primitive n-th roots of unit. Explicitly

Y 2πik Φn(X) = (X − ζk) , ζk = e n . 1≤k≤n k⊥n Here k ⊥ n means that m and n are coprime or relatively prime (a useful notation introduced in [GKP94, section 4.5]), i.e. k ⊥ n ⇔ gcd(k, n) = 1 arXiv:2010.02668v2 [math.NT] 18 Oct 2020 The cyclotomic polynomials are a well-known object in number theory and they also play a key role in ﬁeld theory, see for instance [DF04, section 13.6]. Nice surveys on the subject of cyclotomic polynomials are [Tha00] and [Ge08]. Also the web page [Wei20] collects some known results and references on them.

2020 Mathematics Subject Classiﬁcation. Primary 11C08, 11A25. Supported by ANPCyT under grant PICT-2018-03017, and by Universidad de Buenos Aires under grant 20020160100002BA. The author is a members of CON- ICET, Argentina. 1 2 PABLO L. DE NAPOLI´

In particular, it is known that the cyclotomic polynomials Φn(X) have integral coeﬃcients, and that many important functions in multiplicative number theory are related to them. For instance, it is clear from their deﬁnition that their degree is given by ϕ(n), Euler’s totient function, which counts the number of integers k in the range 1 ≤ k ≤ n that are coprime with n. Also it is easily seen that the cyclotomic polynomials satisfy the following Fundamental identity

n Y (1) X − 1 = Φd(X) d|n (since every n-th root of the unit is a d-th primitive root for exactly one d dividing n), from where we deduce that the cyclotomic polynomials can be computed recursively using the formula Xn − 1 (2) Φn(X) = Q . d|n,d

Moreover, if we consider the coeﬃcients ak(n) of Φn(X), i.e. we write ϕ(n) X k (3) Φn(X) = ak(n)X , k=0 we have that

a1(n) = aϕ(n)−1(n) = −µ(n) for n > 1 where µ is the M¨obiusfunction  1 if n = 1  k µ(n) = (−1) if n = p1p2 . . . pk for distinct primes pj.  0 otherwise Indeed, it is well known that µ(n) gives the sum of the n-th primitive roots of the unit so that

aϕ(n)−1(n) = −µ(n) and a1(n) = aϕ(n)−1(n) by the symmetry of the cyclotomic polynomial [Tha00, Lemma 2.1]. Also from the Fundamental Identity (1) we can derive the expression

Y n/d µ(d) Y d µ(n/d) Φn(X) = (X − 1) = (X − 1) d|n d|n using the multiplicative version of M¨obiusinversion formula (see Lemma 2.2 below). AN IDENTITY INVOLVING THE CYCLOTOMIC POLYNOMIALS 3

Many of the arithmetical functions in multiplicative number theory f : N → R (where R is some commutative ring, usually the ﬁeld C of complex numbers) are multiplicative in the sense that f(m · n) = f(m) · f(n) whenever m ⊥ n For instance ϕ and µ have this property (see [HW79] or [Apo76, chapter 2], and section 3 below). Another arithmetical function closely related to the cyclotomic polynomials is the Ramanujan sum cn(q) (introduced in [Ram00]), deﬁned as the sum of the q-powers of the n-th primitive roots of the unit X q 2πik n (4) cn(q) = ζk , ζk = e . 1≤k≤n k⊥n As we have mentioned before

(5) µ(n) = cn(1) for all n ∈ N. The Ramanujan sums are multiplicative as a function of n (see [HW79, Theorem 67])

cmn(q) = cm(q) · cn(q) whenever m ⊥ n and also satisfy the following more complex multiplicative property [Apo76, Theorem 8.7]

cmn(ab) = cm(a) · cn(b) whenever a ⊥ n and b ⊥ m. Likewise, other arithmetical functions deﬁned by sums involving the roots of the unit, like Gauss quadratic sums and Kloosterman sums, enjoy similar multiplicative properties [HW79, section 5.6].

A question that naturally arises is whether Φn(X), considered as an arithmetical function of n ∈ N into the ring Z[X] of polynomials with integral coefficients, has some property of this kind. In this note, we present an elementary identity involving the cyclotomic polynomials, answering this question. In the first version of this article uploaded to the Arxiv, it is said that this result seemed to be new. However, after that, I have learned that this result has previously appeared in [CMW95, corollary 2](with a different proof). Theorem 1.1. Let m and n be coprime. Then, m Y (6) Φn(X ) = Φd·n(X) d|m 4 PABLO L. DE NAPOLI´

Proof. The proof uses (complete) induction on n . We will show that for each n ∈ N, (6) holds for every m ∈ N such that m ⊥ n.

Indeed for n = 1, Φ1(X) = X−1 and (6) reduces to the Fundamental Identity (1). Next, we assume then that (6) holds for any n0 < n in place of n, and we will show that it holds for n. From (2) (substituting Xm for X), we have that (Xm)n − 1 Φ (Xm) = , n Q Φ (Xm) d2|n,d2

Here we have used the fact that since d2|n, d2 is also coprime with m. On the other hand, from the Fundamental Identity (1),

m n mn Y (X ) − 1 = X − 1 = Φd(X). d|mn Now we observe that the Fundamental Theorem of Arithmetic implies that, since m and n are coprime, each divisor d of mn can be uniquely decomposed as

d = d1 · d2 where d1|m and d2|n. Hence, we can write

m n Y Y (X ) − 1 = Φd1·d2 (X)

d1|m d2|n     Y Y Y =  Φd1·d2 (X) ·  Φd1·n(X)

d1|m d2|n,d2

(splitting the factor with d2 = n). Replacing in (7), it follows that

m Y Φn(X ) = Φd1·n(X)

d1|m as claimed. By the principle of (complete) mathematical induction it follows that the theorem holds for every n, m ∈ N. AN IDENTITY INVOLVING THE CYCLOTOMIC POLYNOMIALS 5

Remark 1.2. It is easily seen that (6) fails if n and m are not coprime. For instance if m = 2 and n = 4 2 m 2 4 Φ4(X) = X + 1 ⇒ Φn(X ) = Φ4(X ) = X + 1, whereas Y 2 4 6 4 2 φd·n(X) = Φ4(X)Φ8(X) = (X + 1)(X + 1) = X + X + X + 1 d|m It is my pleasure to acknowledge that the software Sagemath [The20] was used to ﬁnd this counterexample and to check many of the identities in this work, and to thank their developers for this wonderful tool.

2. A dual form of the main identity In this section, we prove a dual form of our main indentity. Theorem 2.1. If n and m are coprime, Y d µ(m/d) Y m/c µ(c) (8) Φnm(X) = Φn(X ) = φn(X ) d|m c|m

For the proof we need a the M¨obiusinversion formula that we state as a lemma (see [HW79, theorems 266 and 267] for a proof).

Lemma 2.2 (M¨obiusinversion formula). Let f, g : N → R be two functions, where R is a commutative ring. i) (Additive form) The relation X g(m) = f(d) for every m ∈ N d|m is equivalent to X m X m f(m) = g(d)µ = g µ(c) for every m ∈ d c N d|m c|m ii) (Multiplicative form) Assume that R is a ﬁeld. Then, the relation Y (9) g(m) = f(d) for every m ∈ N d|m is equivalent to Y µ(m/d) Y µ(c) (10) f(m) = g(d) = g(m/c) for every m ∈ N d|m c|m Here we make the convention that x0 = 1 even if x = 0. 6 PABLO L. DE NAPOLI´

In our application of the multiplicative form of Möbiusinversion formula, R = Q(x) is the field of rational functions with rational coefficients. Now we see that using the lemma, Theorem 2.1 follows from Theorem 1.1 by fixing n and considering Φ (x) if d ⊥ n f(d) = dn 0 otherwise

Φ (Xm) if m ⊥ n g(m) = n 0 otherwise The relation (9) is just (6) if n ⊥ m. Likewise (10) reduces to (8) when m ⊥ n as d|m implies that d ⊥ n. If not, both sides of (9) vanish as d = m is one of the divisors in the right hand side. Some known properties of the cyclotomic polynomial follow easily from our identity. Corollary 2.3. [Ge08, Corollary 2.3] If p is a prime and k ≥ 1 then,

( pk Φn(X ) if p divides n k pk Φp ·n(X) = Φn(X ) pk−1 if p does not divide n. Φn(X ) Proof. We ﬁrst consider the case in which p does not divide n. We use theorem 2.1 with m = pk.

k pk Y m/c µ(c) Y m/pj µ(pj ) Φn(X ) Φ k (X) = Φ (X ) = Φ (X ) = p n n n pk−1 Φn(X ) c|pk j=0 since by the deﬁnition of the M¨obiusfunction   1 for j = 0 µ(pj) = −1 for j = 1 ,  0 for j ≥ 2 this proves the corollary in this case. If p divides n, we write n = pj · n0 where p does not divide n0. Then, using what we have already proved,

pk+j Φn0 (X ) Φ k (X) = Φ k+j 0 (X) = . p ·n p ·n pk+j−1 Φn0 (X ) Likewise pk Φn0 (X ) Φ (X) = Φ k 0 (X) = . n p ·n pk−1 Φn0 (X ) AN IDENTITY INVOLVING THE CYCLOTOMIC POLYNOMIALS 7

Then, substituting Xpj for X, pj pk pk+j pj Φn0 (X ) ) Φn0 ((X ) Φ (X ) = = = Φ k (X) n pj pk−1 pk+j−1 p ·n Φn0 ((X ) ) Φn0 (X ) as we have claimed. 3. The multiplicative property of Euler’s totient function In this section, we show how identity (6) is related to the multiplicative property of ϕ. We remark that comparing the degree of both sides in the Fun- damental Identity (1) gives a well-known property of Euler’s totient function X (11) ϕ(d) = n. d|m Likewise if we compare the degree of both sides in (6), we get that X (12) ϕ(dn) = mϕ(n) when m ⊥ n. d|m Theorem 3.1. The identity (12) is equivalent to the multiplicative property of ϕ (13) ϕ(mn) = ϕ(m)ϕ(n) when m ⊥ n in the sense that each property can be deduced from the other using (11). Proof. Assume ﬁrst that that ϕ is multiplicative. Then (12) follows easily from (11) since d|m ⇒ d ⊥ n. Therefore, X X X ϕ(dn) = ϕ(d)ϕ(n) = ϕ(n) ϕ(d) = mϕ(n). d|m d|m d|m On the other hand, assume that (12) holds. We will show (13) holds by induction on m (for every n coprime with m). For m = 1, it holds trivially since ϕ(1) = 1. Assume then (13) holds for any m0 < m. Then using (12) X ϕ(dn) + ϕ(nm) = mϕ(n) d|m,d

4. Ramanujan sums In this section, we will apply (6) to the Ramanujan sums (4), and deduce a formula for computing the coeﬃcients of the cyclotomic polynomials. We will make use of the logarithmic derivative operator P 0 L[P ] = P on polynomials. We observe that it has the fundamental property (14) L[P · Q] = L[P ] + L[Q] We will also use the method of generating functions. We need the following lemma (taken from [RPT60, appendix III to chapter X]):

Lemma 4.1. Let P ∈ C[X] be a polynomial of degree N with complex coeﬃcients, N X j P (z) = ajz with an 6= 0. j=0

Let ρ1, ρ2, . . . , ρN be the roots of P (repeated according to their multi- plicity) and let q q q Sq = Sq[P ] := ρ1 + ρ2 + ... + ρN be the sum of its q-powers. Then L[P ] has the following Laurent expansion

0 ∞ P (z) X Sq (15) L[P ](z) = = P (z) zq+1 q=0 for |z| > M = max1≤j≤N |ρj|. AN IDENTITY INVOLVING THE CYCLOTOMIC POLYNOMIALS 9

Proof. We have that

P = an (z − ρ1)(z − ρ2) ... (z − ρN ) Using (14), we have that N X 1 L[P ](z) = z − ρ j=1 j The lemma follows by expanding each term in a geometric series ∞ ∞ q q 1 1 1 1 X ρj X ρj = · = = for |z| > |ρ | z − ρ z 1 − (ρ /z) z z zq+1 j j j q=0 q=0 and adding the results (which is legitimate for |z| > M by the absolute convergence of the series).

By applying this lemma to the cyclotomic polynomial Φn(z) we im- mediately get Corollary 4.2. We have the following Laurent expansion for the logarithmic derivative of the cyclotomic polynomials: 0 ∞ Φ (z) X cn(q) L[Φ ](z) = n = for |z| > 1. n Φ (z) zq+1 n q=0 Remark 4.3. Let P be a polynomial and let Q(z) = P (zm). Then L[Q](z) = m zm−1 L[P ](zm). We are ready to see how property (6) applies to the Ramanujan sums: Proposition 4.4. (6) implies that if n ⊥ m, X m · c (q/m) if m|q c (q) = n dn 0 otherwise d|m Proof. We consider the identity in Theorem 1.1. By taking the logarithmic derivative on both sides and using the previous remark, we get for |z| > 1, m−1 m X m z L[Φn](z ) = L[Φd·n](z). d|m We expand each side in a Laurent series

∞ ∞ ∞   X m cn(r) X X cdn(q) X X 1 = = c (q) . z(r+1)m−(m−1) zq+1  dn  zq+1 r=0 d|m q=0 q=0 d|m 10 PABLO L. DE NAPOLI´

By the uniqueness of the Laurent expansion, X cdn(q) = m · cn(r) d|m when (r + 1)m − (m − 1) = q + 1 ⇔ rm = q, and that the sum is zero otherwise.

Remark 4.5. When q = 0 this property reduces to (12), since cn(0) = ϕ(n). As before, using the additive version of M¨obuisinversion formula, we get Corollary 4.6. If n ⊥ m, X q m c (q) = d c µ . mn n d d d| gcd(m,q)

In particular, if we choose n = 1, c1(q/d) = 1 and we get the following known explicit formula for the Ramanujan sums due to Kluyver [Klu06] See also [Apo76, Theorem 8.6] X m (16) c (q) = d µ ∀m ∈ . m d N d| gcd(m,q) Another explicit formula for the Ramanujan sums is

m µ gcd(m,q) ϕ(m) (17) cm(q) = m ϕ gcd(m,q) The function on the right hand side was initially studied by von Ster- neck [Ste02]. Later, Kluyver [Klu06] and also Hölder[Höl36]proved that it coincides with the Ramanujan sums. See [FGK14] for more information on the Ramanujan sums and their history. We conclude this note by explaining how the coefficients of the cyclotomic polynomial Φn(X) can be recursively computed using the Ra- manujan sums, without the need of factoring polynomials.

Lemma 4.7 (Newton Relations). Let P ∈ C[X] of degree N and consider the sums Sq of the q-powers of its roots as in lemma 4.1. Then the coeﬃcients aj of P are related to the sums Sq by:

`−1 1 X a = − a · S for j = 1, 1, 2,...,N − 1 N−` ` N−j `−j j=0 REFERENCES 11

This result follows from Lemma 4.1 by writting (15) as N N ! ∞ ! X X X Sq r a zr−1 = a zj · r j zq+1 r=1 j=0 q=0 and equating the coeﬃcients on both sides. See [RPT60, appendix III to chapter X] for details1.

Corollary 4.8. Let Φn(X) be the cyclotomic polynomial. Its coefficients aj(n) (for 0 ≤ j ≤ N = ϕ(n)) can be recursively computed in terms of the Ramanujan sums using the relation `−1 1 X a (n) = − a (n) · c (` − j) for ` = 1, 2,...,N − 1, N−` ` N−j n j=0 starting from aN = 1 Together with (16) or (17) these formulas provide an algorithm for computing Φn(X) without the need of dividing polynomials. More information on the coefficients of cyclotomic polynomials and their relations to other arithmetical functions can be found in [HM20]. We also refer those readers who are interested in efficient algorithms for the computation of cyclotomic polymonials to [AM11] and [Bre93].

References [Apo76] Tom M. Apostol. Introduction to analytic number theory. Undergraduate Texts in Mathematics. Springer-Verlag, New York-Heidelberg, 1976, pp. xii+338. [AM11] Andrew Arnold and Michael Monagan. “Calculating cyclotomic polynomials”. In: Math. Comp. 80.276 (2011), pp. 2359– 2379. issn: 0025-5718. doi: 10.1090/S0025-5718-2011- 02467-1. url: https://doi.org/10.1090/S0025-5718- 2011-02467-1. [Bre93] Richard P. Brent. “On computing factors of cyclotomic polynomials”. In: Math. Comp. 61.203 (1993), pp. 131–149. issn: 0025-5718. doi: 10.2307/2152941. url: https:// doi.org/10.2307/2152941.

1 j Beware that in this book the notation for the coeﬃcient of X in P is aN−j instead of aj. 12 REFERENCES

[CMW95] Charles Ching-an Cheng, James H. McKay, and Stuart Sui Sheng Wang. “Resultants of cyclotomic polynomials”. In: Proc. Amer. Math. Soc. 123.4 (1995), pp. 1053–1059. issn: 0002-9939. doi: 10.2307/2160701. url: https://doi. org/10.2307/2160701. [DF04] David S. Dummit and Richard M. Foote. Abstract alge- bra. Third. John Wiley & Sons, Inc., Hoboken, NJ, 2004, pp. xii+932. isbn: 0-471-43334-9. [FGK14] Christopher F. Fowler, Stephan Ramon Garcia, and Gizem Karaali. “Ramanujan sums as supercharacters”. In: Ra- manujan J. 35.2 (2014), pp. 205–241. issn: 1382-4090. doi: 10.1007/s11139-013-9478-y. url: https://doi.org/ 10.1007/s11139-013-9478-y. [Ge08] Yimin Ge. “Elementary properties of cyclotomic polynomials”. In: Mathematical Reflections 2 (2008). [GKP94] Ronald L. Graham, Donald E. Knuth, and Oren Patash- nik. Concrete mathematics. Second. A foundation for com- puter science. Addison-Wesley Publishing Company, Read- ing, MA, 1994, pp. xiv+657. isbn: 0-201-55802-5. [HW79] G. H. Hardy and E. M. Wright. An introduction to the theory of numbers. Fifth. The Clarendon Press, Oxford Uni- versity Press, New York, 1979, pp. xvi+426. [HM20] Andrés Herrera-Poyatos and Pieter Moree. “Coefficients and higher order derivatives of cyclotomic polynomials: old and new”. In: Expositiones Mathematicae (2020). [Höl36] Otto Hölder.“Zur Theorie der Kreisteilungsgleichung K m(x) = 0”. In: Prace Matematyczno-Fizyczne 1.43 (1936), pp. 13– 23. [Klu06] JC Kluyver. “Some formulae concerning the integers less than n and prime to n”. In: Proc. R. Neth. Acad. Arts Sci.(KNAW) 9 (1906), pp. 408–414. [Ram00] S. Ramanujan. “On certain trigonometrical sums and their applications in the theory of numbers [Trans. Cambridge Philos. Soc. 22 (1918), no. 13, 259–276]”. In: Collected pa- pers of Srinivasa Ramanujan. AMS Chelsea Publ., Provi- dence, RI, 2000, pp. 179–199. doi: 10.1016/s0164-1212(00) 00033 - 9. url: https : / / doi . org / 10 . 1016 / s0164 - 1212(00)00033-9. [RPT60] Julio Rey Pastor, Pedro Pi Calleja, and CésarA. Trejo. Análisismatemático. Vol. I: Análisisalgebraico. Teoria de ecuaciones. Cálculoinfinitesimal de una variable. Fifth edi- tion]. Editorial Kapelusz, Buenos Aires, 1960, pp. xxviii+836. REFERENCES 13

[The20] The Sage Developers. SageMath, the Sage Mathematics Soft- ware System (Version 9.1). 2020. url: https : / / www . sagemath.org. [Ste02] RD von Sterneck. “Sitzungsber”. In: Math.-Natur. Kl. Kaiserl. Akad. Wiss. Wien 111.1567-1601 (1902), pp. 1–14. [Tha00] Ravindranathan Thangadurai. “On the coeﬃcients of cyclotomic polynomials”. In: Cyclotomic ﬁelds and related topics (Pune, 1999) (2000), pp. 311–322. [Wei20] Eric W Weisstein. “Cyclotomic Polynomial”. In: MathWorld– A Wolfram Web Resource (2020). url: https://mathwor ld.wolfram.com/CyclotomicPolynomial.html.

IMAS (UBA-CONICET) and Departamento de Matematica,´ Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Ciudad Universitaria, 1428 Buenos Aires, Argentina Email address: [email protected]